Kolibri Load‑Testing CLI — Quick Start

Run from integration_testing/load_testing/; entrypoint is loadtest.py.

Prerequisites

Before running load tests:

  1. Install dependencies:

    pip install -r requirements/load_test.txt

  2. Start Kolibri server (NOT development server):

    # Use port 8080 for production-like server
kolibri start --port=8080

    Server state: The server can be either:

    • New KOLIBRI_HOME/unprovisioned - the tool will provision it for you

    • Existing KOLIBRI_HOME - the tool will use existing setup

    Important: Do NOT use pnpm run devserver - load tests must run against a production-like server to get accurate performance measurements. The development server has additional overhead that will skew results.

Help

python loadtest.py --help
python loadtest.py --help
python loadtest.py run --help

See the help for available flags (e.g., users, spawn rate, duration, headless, retries).

Full provision → manual capture → test

Runs the whole flow against a Kolibri server (provisioned or unprovisioned):

  • Fresh server: Provisions device → creates facility/users/class → imports QA channel → creates lesson

  • Existing server: Uses existing setup where possible, creates missing components

The tool prompts for admin credentials which are used to either:

  • Create a NEW admin account (if provisioning a fresh server)

  • Authenticate with EXISTING admin account (if server already provisioned)

Then opens a browser for manual capture → saves a versioned HAR → executes the Locust test.

python loadtest.py

Full run using an existing HAR (skip capture)

Use a pre-recorded HAR to skip the manual capture step. With a file in har_files/lesson_flow_kolibri_testing_high_latency.har, for example:

python loadtest.py --har har_files/lesson_flow_kolibri_testing_high_latency.har

Other flags

Non-interactive server and credentials:

python loadtest.py --server http://127.0.0.1:8080 --username admin --password sosecure

Headless 10‑minute run at modest scale:

python loadtest.py --headless -u 100 -r 50 -t 10m run

HAR Files

HAR files capture the exact sequence of HTTP requests made during a learner lesson interaction.

When to use existing HAR files:

  • The committed HAR file’s version is ≥ your Kolibri version AND there have been no frontend request pattern changes since that version

  • Example: lesson_flow_kolibri_0.18.4.har can be used for testing 0.18.4, 0.18.5, 0.18.6, etc., as long as no frontend changes affected learner request patterns

When to create new HAR files:

  • Frontend code changes that modify the pattern of HTTP requests during learner interactions

  • New Kolibri version where request patterns have changed

  • Adding new content types or interaction flows to test

Versioning: HAR files are named with the Kolibri version where they were captured (e.g., lesson_flow_kolibri_0.18.4.har). This version acts as a “valid from” marker - the HAR can be used for that version and later versions until request patterns change. Committed HAR files for released versions are preserved in version control.